See the Forest and the Trees with Bungee View

Bungee View is a visualization prototype developed at Carnegie Mellon University to support casual users gaining an understanding of an image collection as a whole, and in finding patterns in such collections. As a research experiment with the University of Pittsburgh, you can explore the Visuals for Foreign Language Instruction image collection hosted by the University Library System's Digital Research Library.

For a brief tour, watch this two minute video or just read its transcript.

Run Bungee View

  1. Click to start Bungee View as a Java Web Start application in a new window, or for help if it doesn't start.
  2. Once it starts, click on any image, bar, or underlined text.
  3. In the figure below, in the blue area at the top, especially notice what's inside the labeled yellow boxes: the Help Menu, the “You Are Here” description of your current query in English, and the description of What Clicking Will Do.
Screenshot showing Help Menu, You Are Here, and What Clicking Will Do.
In this screenshot, the heights of the Date bars from the 17th century to the 20th century show that the proportion of works in the collection from North America increases over this period.

Data-Mining with Bungee View

Traditional web search is like finding trees in a forest. Data mining is like finding Tags describing patterns of trees. Tag Clouds are one way to visualize such patterns in text collections.

Traditionally, data mining has used structured data, like that found in library card catalogs, rather than raw text. Bungee View uses structured data, and shows the patterns and the trees. Rather than Tag Clouds, it uses two other overview visualizations, Top Tags and Tag Walls:

Image of a Tag Cloud and Bungee View's alternative visualizations

Tag Cloud

Top Tags

Tag Wall

Quick summary of a set of documents

More quantitative summary

Richer summary

There is no meta-data that “20th Century” is a single concept, which lessens the potential to support search

Supports search

Supports search

Tag importance is shown by font size and color

Tag importance is shown by order, and numerically

Tag importance is shown by rectangle size and shape

Up to ~30 characteristic tags (uncharacteristic tags could be shown using color)

Up to ~14 characteristic and uncharacteristic tags

Up to ~300 characteristic and uncharacteristic tags (and even more when you zoom or filter)

Summarizes Collection: rectangle width shows that more than half of the works in the collection are from the 20th Century

Compares Subset to Whole: rectangle height upward and green color shows that the current results include an even higher percentage of 20th Century works than the collection as a whole

Organizes tags by category, like Date and Format

Systematic and consistent tag layout creates stable “maps” of the collection

With hundreds of organized and sorted tags to filter on, you may not even need the text search box. This is important in image collections with little text, since the keywords you have in mind may not occur at all.


Web search engines are widely used for information retrieval from unstructured text. The amount of structured and semi-structured information available on the Web is also huge, and is more amenable to data mining. Yet there has been no similar explosion of interest in this kind of data. Finding patterns in databases of political contributions, environmental data, or hospital and school performance would surely interest many citizens. The main research question for this project is how to support such exploration for those with little or no training in statistics or programming. In contrast to other data-mining systems, Bungee View focuses on learnability, responsiveness, and providing a satisfying experience. It has so far been used on image collections because the “trees” are easy to visualize compactly, while text-based systems have little to work with.

Have a collection?

Let me help you make it available through Bungee View! The project site contains the source code and documentation for Bungee View Administrators and Developers.


Since this is a research project, your usage will be recorded to help improve Bungee View. We will use your IP address to track your repeat visits, but will not attempt to identify you as a person.

Feedback is encouraged.
Last update: 10 October 2014
Mark Derthick